hot topics

I'd rather hold off on starting the individual player threads. I liked the one-a-week approach. No reason we can't argue about Larry Walker here for awhile.

I'd rather hold off on starting the individual player threads. I liked the one-a-week approach. No reason we can't argue about Larry Walker here for awhile.

Definitely, Devin. Besides, posting too many threads at one time will make them more difficult to find under Hot Topics.

Is there anyone to work up MLE's for Nomo? Are we allowed to consider Japanese stats?

Not that it'll matter in his case, but, hey, why not take a look.

Just eyeballing this and pondering my consideration set, I'm pretty sure Bagwell's going to push McGriff off my ballot. Walker will probably push Cravath off my ballot (if not, Cravath will be right down the bottom). Brown might push Walters off my ballot.

I'm actually quite interested to see the Larry Walker discussion, if only because I'm looking forward to see how different people handle concerns over any distorting effects of Coors on his stats over and above the normal park factors, the impact and relative evalautions of his defense, and considerations relating to his in-season durability.

Plus, the HoM is currently woefully lacking in Canadian content.

I disagree.

I'd rather see the Walker evaluation on it's own thread so it doesn't get lost.

This should be the thread for provisional ballots, ranking players amongst each other, evaluating the backlog, etc..

But items specific to a player - items that 2 years from now you are going to say, 'how good was Walker's baserunning and defense?' - those should be for reference on the player's threads.

I agree we don't want to splinter the discussion too much, but do you see what I'm saying?

Why is Hampton there?

Why is Hampton there?

John forgot to delete him.

Here's how I currently have the new candidates ranked, method is by adding up each season where the player was above average (wins above average) and ignoring all other seasons -so a player isn't hurt by a terrible year or two (which helps Nomo)- but isn't helped by a long list of average-ish years (Grissom/Surhoff)



Hampton if eligible would slot in around Franco
Nomo's MLB score is 9.2, he is ranked above after being given some half-assed credit for Japan.

What a class! I don't have access to my ballot spreadsheet, but I sketched up my 2011 prelim last night. From memory, it's pretty close to this:

1. Bagwell
2. L. Walker
3. Cone
4. K. Brown
5. Easter

6. B. Bell
7. Palmeiro
8. Bancroft
9. Williamson
10. Reuschel

11. Appier
12. Willis
13. Dunlap
14. Mullane
15. Olerud

Is there anyone to work up MLE's for Nomo? Are we allowed to consider Japanese stats?

Yes to the second. Japanese players who played in North America are considered for their whole career including their time spent in Japan. Similarly, North American players who spent some time in Japan can have those seasons considered as a different kind of minor league credit.

Plus, the HoM is currently woefully lacking in Canadian content.

Tip O'Neill for the Hall of Merit, eh!

Hugh Duffy made it into the top 10, and hence requires comment. (I had him on my 2010 ballot; his placement on my 2011 ballot is less certain.)

The fact that he's in the top 10 - 7th, in fact - feels almost accidental. I tallied incrementally as votes came in, and every time I added a now vote or two, the lower part of the top 15 would shuffle around. Cone sticks up a bit above the backlog, but from Rizzuto (5th place, 220 points to Concepcion (15th place, 163 points) the differences are tiny.

I've always supported 1890's outfielders, and my order has always been (and still is) Van Haltren > Duffy > Ryan. Van Haltren is one of the very, very few players I've ever had in an "elect me" slot who hasn't been elected. But going forward?

These three are the third string.

The first string of 1890's-centered outfielders was Burkett, Hamilton, and Delahanty. They sailed into the HoM easily, as they should have - I don't think any of us have any qualms about them.

The second string was Keeler and Kelley. They also got in pretty easily, but if you scratch the electorate hard enough, you might find a qualm or two.

That makes Van Haltren, Ryan, and Duffy the third string. They started out in the upper backlog and never made it over the hump; there are instances in which candidates who started out behind them surged past them to election. Ryan has even periodically disappeared from the ballot even though he's objectively hard to distinguish from the other two. And the judgment that Duffy should be a much stronger candidate than Van Haltren - that wasn't so obvious to the electorate back when they were newer on the ballot and there were fewer distractions.

One particular comment about Duffy's 1894 offensive year: it was a very good year, of course. But after you deflate the 1894 out of it, it doesn't look earth-shattering to me. It wasn't anything close to George Stone's 1906. (Not that Stone, with his microscopically short career, is any kind of candidate.)

Take this first string/second string/third string idea and apply it to 1980's-1990's outfielders. (One problem: it matters more now who is and isn't a CF than it did in 1895.)

The first string is Bonds, Griffey, and Henderson.

The second string is Raines and Gwynn (no CF).

(And I'd rather have Bonds/Griffey/Henderson than Burkett/Hamilton/Delahanty, and I'd rather have Raines/Gwynn than Keeler/Kelley.)

So who's the third string? Dawson, Puckett, Winfield? And Ramirez, Belle, Walker? Only now I realize that I've cast the temporal net pretty wide; Walker and Ramirez aren't really contemporaries of Dawson and Winfield. That's wider than the times included in the 1890's-centered talk. More time, more teams, more eligible players (including that 11 of the 11 players I just named have too much melanin in their skin to have been allowed to compete with the 1890's guys.)

Looking at Baseball Reference pages.... where did league average ERA go? I need that as an input for my pitchers!

Looking at Baseball Reference pages.... where did league average ERA go? I need that as an input for my pitchers!

You could reverse-engineer a good estimate for it from ERA+: (pitcher ERA) * ERA+ / 100.

Sure, but I'm concerned about the missing decimals screwing up the calculation.

John forgot to delete him.

I didn't forget, since you mentioned updating the list and I was waiting for that, Dan. With that said, I'll remove him in a few minutes.

James Newburg, would you care to enlighten us on Cone over Brown??? Again...3250 IP with a 127 ERA+ kinda schools 2900 and 120, dunnit??

James Newburg, would you care to enlighten us on Cone over Brown??? Again...3250 IP with a 127 ERA+ kinda schools 2900 and 120, dunnit??

I don't know about James, but as a New Yorker, I have to say Cone is vastly superior in every way to Brown... my spreadsheet just doesn't see it though:-)


Cone's best 6 (ops+/ip)
170/171 (strike year)
159/195
146/231
138/254
137/193
131/229 (post strike lead league in IP)

Brown
216/233
169/230
169/211
164/257
150/237
143/252

So in their better years Brown both had better rate stats and pitched more innings

Sometime in the next year I'm going to look at incorporating DanR's and Chone's WAR into my ratings. Right now I have the newbies:

Bagwell
Palmeiro
Brown

1-2-3 on my prelim

Olerud (just below McGriff)
Walker (right around Norm Cash)

Here's my version of JPWF13's post, including Tiant, whom I voted for in the recent election. These are RA+ equivalent records. Since it's RA, not ERA, if Brown allowed more than his share of UER, he's being charged with those. The years are sorted by single-year Fibonacci win points.

Walker right around Norm Cash?? OK, time for Dan to go apoplectic. (I know, I know, the vote's a year away). OK, as hitters, they're even--they both have 8,000 PA of 140 OPS+. Ah, but wait, you say...there's more to baseball than just hitting? Oh, right. Walker

a. was an outstanding baserunner (3 wins)
b. played a more valuable position (9 wins), and
c. played it brilliantly (11 wins).

If you add up Walker's baserunning, positional, and fielding value (which are backed strongly by anecdotal/reputational accounts as well as the play-by-play statistics available for the modern game), and transfer it to his bat to make him the equivalent of an average defensive first baseman, his career OPS+ would increase by 21 points to 161. That puts him right around Mark McGwire, who sailed into the HoM (albeit with a slightly higher peak).

James Newburg, would you care to enlighten us on Cone over Brown??? Again...3250 IP with a 127 ERA+ kinda schools 2900 and 120, dunnit??

Not necessarily. I've seen bigger ERA+ difference trumped by team defense, unearned runs and some other things (bullpen support, leverage when relieving, quality of league relative to the other league that year), for one. I haven't looked at Brown/Cone yet, so I have no idea if that applies.

I see ERA+ as much more 'shaky' metric than say OPS+ for example. Even though OPS+ doesn't account for baserunning, GIDP, etc. (meaning I know OPS+ has it's issues).

I've had some pretty large differences (see Palmer, Jim) between my DRA+ and ERA+ once you start looking deeper.

BTW, just in case it wasn't clear; the moral of #23 isn't the Cone > Brown, it's ERA+ is an decent eyeball metric at best. It is nowhere near a be all end all. You absolutely need too look a lot deeper than ERA+.

OK, time for Dan to go apoplectic.

Heh, I knew that would get you. Anyway, I said I was going to work on adding your data, so we'll see what happens.

What do you mean by "bullpen support"--the pct of inherited runners left by a SP that score? Do you have the data for that?

I am going to put this as a placeholder for when I come back to look at this at the end of next year. My first glance has Bagwell 1, Palmeiro 2, Brown 9 (behind Joss, ahead of Tiant) and Walker 10. I'll be reading Dan R's impassioned pleas for Walker (but you can stop shouting, some us are Canadian) and trying to figure out how much credit to give him for his fielding and defence but his career is pretty short for a a career oriented voter like me.

Olerud will be off the ballot. I think the stats back up the perception, but as a Blue Jay fan if you were going to ask me who seemed to have the HoM career -- it was the Crime Dog.

Bringing this over from the 2010 results thread, with a few updates;

Very Preliminary Rankings for 2011:

1. Bagwell 175 (no one's close!)
2 Reuschel 145
3. Brown 137 (a little lower than I expected, but this is just a preliminary number)
4. Walker 134 (just ahead of the next one on intangibles)
5. Palmeiro 134.4
6. Cravath 129.3
7. Campaneris 129.4
8. Rizzuto 127.8
9. Concepcion 127.7
10. Tiant 126.6
11. John 125.7
12. Dunlap 125.1
13. Grimes 125.5
14. Rucker125.2
15. B. Bell 122.2

It's a strong entering class. It's clear there's been some movement on new metrics, and the extension back in time of better base-running and fielding metrics, which I'll need to account for in firming up the order for the backlog.

Olerud is next among the newcomers. He checks in at 102, which is well outside the top 50.

What do you mean by "bullpen support"--the pct of inherited runners left by a SP that score? Do you have the data for that?

Yes, all the way back to 1957 or 1960 I think (it's been a little while since I've entered anyone new).

If you are a BPro subscriber, you can download the data for every year to CSV files. And it's already been converted to runs by them.

I've tried to account for as much as possible with my pitcher ratings. I was really excited when I found that in such an easily usable format.

It also works on the flip-side, as it has inherited runner data for relievers too. That's one of the things that makes John Hiller's 1973 (and Goose Gossage's 1977) so insanely good. Those two years were up there with Koufax's best in terms of value, and I would never have known that without that data.

B-R has bequeathed runner data as well (I don't know if BPro's has more information).

Cone left on 203 runners (0.070 per IP) and 72 scored (35.5%).

Brown left on 247 (.076/IP) and 92 scored (37.2%).

One would assume that over such long careers, there would be no real difference in the types of runner bequeathed (average base/out state), but who knows if that's really true.

As a quick example of the first guy I noticed with some large numbers . . . Jim Kaat's relievers saved him 10.6 runs when he was a SP over his career. In 1965 alone they saved him 4.9.

As a reliever, he cost the guys he relieved 15 runs over his career. While he was relieving, the guys who came in after him saved him another 13 runs.

Over the course of his career, that's a total of 38.6 runs that he was 'given'. That's .10 off of his R/9 or about 2 points of ERA+ just from inherited runners analysis.

Dick Drago is another one (only a few rows up from Kaat). He looks 27 runs better than he actually was, and that's in just 1875 IP for his career.

Warren Spahn was saved 11 runs in just the 1960-65 period. I've incorporated 1960-present in my numbers, but now I think they go back to 1957 or maybe a little further.

Take Hiller's 1973. He allowed 21 runs in 125 1/3 IP. But he prevented 15.6 inherited runs from scoring, and his relievers cost him 7/10 of a run. You factor that in, and he was as effective as someone that gave up 6.1 runs in 125 1/3 high leverage (1.72) innings. His defense's cost him another .08 on the ERA to boot.

It's very important to take this into account in the era of the relief ace, not quite as big of a deal in the age of the closer. Rivera's entire career (1995-2006) his range was -4.9 to 2.8 and his relievers' range was -1.1 to 1.6, and for his career the totals are he cost his team a run and his relievers saved him 3.

But for some guys it can make a big difference, especially seasonally.

Again, there are two separate preferences here: peak vs. career, and where you set replacement level. E.g.:

A peak voter with a high replacement level (someone who looks at, say, top 3 or 5 seasons measured by runs above positional average) will be attracted to guys with very high rate stats in short careers who may have had durability issues. John McGraw and Frank Chance leap to mind. Walker's rates were excellent, but not otherworldly like McGraw's.

A peak voter with a low replacement level (someone who looks at, say, top 3 or 5 seasons measured by raw Win Shares) will prefer guys with short careers but who could stay on the field in-season. Al Rosen and Albert Belle would be good examples. Walker missed too many games to get much of a look from these voters.

A career voter with a high replacement level (someone who looks at, say, career runs above position average, discarding all below-average seasons) will be attracted to guys with high rate stats over decently long careers who may have had durability issues in-season. This is, I think, Walker's natural constituency. The recently elected Larkin is another case, as is Willie McCovey.

A career voter with a low replacement level (someone who looks at, say, raw career Win Shares) will prefer guys who played forever regardless of how good they were. Think Rusty Staub, Atanasio Pérez, or Rabbit Maranville.

Sam McDowell goes the other way. His relievers cost him 19 1/2 runs. For his career he ends up 410 runs above replacement - that's a little less than 5% of his total career value.

Drysdale is another - horrible bullpen support, cost him 16 runs. As a reliever he saved his guys another 6 runs. And this only counts after 1960. Nolan Ryan was cost 16 runs. Whitey Ford lost 18.

On the flip side, Koufax was saved 20 runs. In a career that was 603 runs above replacement, you'd be overrating him by 3% if you didn't take this into account.

Again, it's not a 'huge' thing, but it definitely is enough to matter on the margins, and can make a huge difference in an individual season, when you are trying to figure out who deserved the Cy Young Award, peak value, etc.

Those are good numbers, Joe D. Since it will be two years before another backlogger gets inducted, I think I'll have plenty of time to study and incorporate them. :)

After the near-perfect 9-man starting lineup that could be crafted from the 2010 eligibles list, this is not nearly so much fun. Gotta play some people out of position. The pitching rotation is much better than that 2010 bunch.

C: Surhoff
1B: Olerud
2B: Boone
3B: Baerga (out of position)
SS: Offerman (yikes!)
LF: Bagwell (out of position)
CF: Grissom
RF: Walker
DH: Palmiero

Reserve C: ??
Reserve IF ??
4th OF: Mondesi
PH: T. Martinez

SP: Brown
SP: Leiter
SP: Nomo
SP: Alvarez
SP: Mulholland
Closer: Franco
RP: Urbina

OK, we've got pitchers, but I don't know how much they're going to like it, with a guy who quit catching forced back behind the plate, a 2B playing 3B, a 1B playing LF, and Jose Offerman at SS.

When you are looking at bullpen support, you can't just look at whether inherited runners score or not; you also have to look at the base/out situation. If the starting pitcher leaves a runner at 3B with no one out, that's a far cry from leaving a runner on 1B with two out. I don't know if BPro actually did this when they calculated bullpen support.

-- MWE

I wouldn't be surprised if there were a few guys who fell a little short of DanG's threshold for listing who might smooth out the positional tangles of the lineup in #36. But there's still going to be no easy way to get all those 1B into the lineup. Bagwell was a least fast enough to steal bases - that's why I thought of him as a possible OF.

Mike, since they convert it to runs, I'm almost positive they did. I cannot remember what I read that lead me to believe that, but I'm almost positive they look at the expected runs when the reliever takes over, vs. how many of the inherited runners he actually let in.

The have to do this right? How else would they have decimal values?

OCF - didn't Palmeiro actually play in the OF for awhile? I'd put him in LF and DH Bagwell.

To John or Joe,

I think that for the future the 201X Ballot Discussion thread should have the 201X-1 Results at the top, not just the new eligibles. Continuity is harder to maintain over the course of a year's time and I think that would help make newbies aware of what all is happening here. The idea of perpetual eligibility might be foreign to a lot of people.

Wow, Chris Cobb, Reuschel over Brown? I'm one of Reuschel's biggest fans, but gosh. Could I trouble you to break that one down?

Wow, Chris Cobb, Reuschel over Brown? I'm one of Reuschel's biggest fans, but gosh. Could I trouble you to break that one down?

Sure. I would emphasize that my analysis of Brown has only been through one step, so the other side of my analysis could move him above Reuschel. I am working with the quick-and-dirty component of my system for pitchers only, since I've been away from the HoM project for awhile and have very little time. But the key factors are these:

First, their career lines from BP's WARP1:

Reuschel: 3548 IP, 3.78 DERA, 81.1 WARP1
Brown: 3256.3 IP, 3.74 DERA, 69.9 WARP1

So for his career, Reuschel has almost 300 IP on Brown at virtually the same level of effectiveness, and he also picks up two wins over Brown with the bat, over his career. That's quite a bit of ground for Brown to make up. Brown undoubtedly has a better peak than Reuschel, but here's how their best five consecutive seasons compare by WARP1:

Reuschel, 1976-90 1250.7 IP (.98 Durability Factor*), 34.1 WARP1, 6.13 W1/225 IP. Peak factor in my system is rate X durability X 5, so Reuschel's peak score is 30.1
Brown, 1996-2000 1209.7 IP (1.03 Durability Factor). 36 WARP1, 6.70 W1/225 IP. Brown's peak score is 34.5

*Durability factor is IP over peak, divided by (1.25 X IP) for an average starter over those five years, which is what I define as an "ace workload." Pitchers who exceed an ace workload during their peaks get a bonus applied to their peak rate; pitchers who fall short of an ace workload take a penalty. The vast majority of pitchers I have studied (pitchers who are marginal to serious HoM candidates) are within 5% of the ace workload during their peak years. The top and bottom outliers are about 20% above and below, respectively.

There's another career above average factor in my system, but since Brown and Reuschel were pretty close career wise in total value above average, it doesn't affect much.

By my system's take on WARP1's numbers, Reuschel's somewhat superior career (as effective as Brown's, but longer), outweighs Brown's somewhat superior (but not greatly superior) peak.

So that's where my numbers come from. Are there numbers here, for Reuschel or for Brown, that look fishy? Are there major adjustments that would seem appropriate?

I will be studying the numbers for both pitchers from other sources and other angles, but this is how my initial take breaks down into its main components.

Lowest-tech stat you'll ever see, but ...

Top 10 in IP

KeBrown 1 2 2 6 7 10

Reuschel 4 7 7 7 8 9

Quick Brown v backlog comparisons out of my spreadsheet:

Pitcher PWAA BWAA2 WARP2 (Dan R)
Kevin Brown 33.3 -7.3 66.2
David Cone. 26.0 -5.1 57.0
Tom Bridges 29.7 -10.2 56.5 (War credited)
Luis Tiant... 20.6 -3.5 53.5
Urb Shocker 24.8 -2.4 51.4 (WWI credited)
Kevi Appier 25.3 -1.6 51.7
Ri Reuschel 24.4 -12.1 60.2

Brown is WAY ABOVE those guys (6-12 WARP).

Similar elected pitchers to Kevin Brown in my spreadsheet - Hal Newhouser, Whitey Ford, Ted Lyons, Bill Foster, Ferguson Jenkins, Ray Brown. I have Brown slightly ahead of those guys for 24th place all-time. This is partly due to his advantage in PWAA. That PWAA number is similar to Roberts and Perry but Kevin Brown didn't have as graceful of a decline (less bulk).

One thing with Brown, Cone, Appier and the others from the 90s . . . Dan has mentioned this before, but if we are normalizing seasonal innings, we also need to normalize for career length. Still not sure how to do/approach this.

If not, we are going to overrate the new guys.

Yes, Joe D, and that does seem to me to be the great challenge. Nonetheless, I don't think there's any doubt that Brown exceeds every possible standard, as do Mussina, Smoltz, and Schilling, and of course their five contemporaries above them. The line is either going to be between those four and Cone, or between Cone and Appier. Fortunately, we'll have two years to work on this before Cone has another shot at getting elected, by which point we can hopefully have some clearer answers/ideas as to how to address this conundrum.

Still not sure how to do/approach [normalizing for career length].

Well, what's the simplest method? Start from there - calculate an average career-length based on a rolling 'n'-year average, and adjust career WARP by that.

Sure, you'll shoot the idea full of holes, but it'll get a discussion going of where the problems lie. Beats muttering about 'oh, there's a problem' while we all hide in our cubicles trying to figure it out.

Fra paolo - good idea, but how do you even calculate an average career length? Who's in the dataset?

What I've been sorta doing is breaking them down into groups by the years they played, and looking at who stands out in each group and where the list sort of 'floods', i.e. where does the 'big bunching' start? Then cut it off just above that. It's an eyeball process.

The problem with that is the same problem you run into with standard deviations--what if you just happen to have a star glut or drought? Did P in the '40s and '80s burn out because it was harder to pitch for a long time then, or just because they weren't very durable? I got around this with stdevs by doing a regression onto league factors like run scoring and expansion, but I don't think that can work for career length where you're basically looking at a group of what 10-15 individuals at any given time?

There's really no use comparing Bagwell to the backlog, so I'll compare him to elected 1B.

player................ BWAA2 BRWAA2 FWAA2 WARP2
Greenberg, Hank 68.0 0.3 5.2 83.3 (WWII credited)
Bagwell, Jeff...... 66.0 2.4 3.9 75.7
Leonard, Buck.... 63.0 0.0 2.0 72.0 (estimated)
Murray, Eddie..... 53.3 -2.6 4.9 66.5
McGwire, Mark.... 58.5 -2.7 0.3 63.8

Bagwell is 10 wins ahead of the recently elected McGwire and Murray partly because of a 5 win advantage on the basepaths. He's behind Greenberg, probably due to his career-ending shoulder injury. Not sure how Bagwell and Greenberg compare on a peak basis. These numbers lead to me ranking Bagwell as the 8th best 1B to date.

Is career length impacted by the in-season innings? If so you'd be adjusting one up and the other down when leaving them alone might be the answer. Can you see TJ surgery in the career length data? If not, then I wouldn't worry about it.

Continuing with Larry Walker

player BWAA2 BRWAA2 FWAA2 WARP2
Walker, Larry 42.7 2.5 10 67.2

Johnson, Bob 48.6 0.5 1.5 60.5 (minor league credited)
Cravath, Gavy 54 -0.8 -2.3 59.4 (minor league credited)
Bonds, Bobby 34.6 2.5 5.5 54
Cuyler, Kiki 34.7 3.1 5.4 52.9

Well, he's way above the backlog. Gavy Cravath has him beat at the plate but gives it all back baserunning and fielding.

Clemente, Roberto 44.3 1.8 12.7 70.2
Jackson, Reggie 60.5 0.0 -1.8 74.9
Slaughter, Enos 50.3 0.5 6.1 71.2 (war credited)
Heilmann, Harry 64.0 0.0 -4.7 70.4
Keeler, Willie 43.5 1.3 11.2 63.6
Flick, Elmer 50.1 1.8 4.3 62.0
Evans, Dwight 37.2 1.1 8.3 66.0

Dwight Evans PLUS is a great way to think of Larry Walker. He comps well to Keeler also.

A few more comparables for Walker - LF

Williams, Billy 51.4 1.8 5.7 69.0
Simmons, Al 44.1 0.6 9.7 66.5
Sheckard, Jimmy 36.1 1.4 13.6 62.2
Wheat, Zack 47.1 0.2 4.3 64.6
Kelley, Joe 46.1 0.4 5.6 63.1

Al Simmons jumps out as another comparable.

Now Raffy Palmeiro

player......... BWAA2 BRWAA2 FWAA WARP2
Palmiero, Raf 49.3 -1.4 4.3 65.8
Taylor, Ben.. 37.0 0.3 8.0 60.0 (estimated)
Cash, Norm.. 47.7 -1.5 6.9 55.4
Olerud, John 35.2 -1.8 10 52.6
Fournier, Jack 49.6 -1.3 -4.1 54.1

He's above the backlog considerably. How about similar elected 1B?

Killebrew, Harmon 60.4 -1.1 -6.8 60.6
Start, Joe........ 35.2 0.0 0.5 71.9
McCovey, Willie. 65.7 -1.9 -6.0 60.8
Hernandez, Keith 43.4 -0.1 9.1 56.6
Beckley, Jake... 37.8 -1.9 3.6 66.2
Clark, Will....... 49.5 -0.3 0.2 55.9

There's a career length advantage on Clark, a hitting/peak advantage on Beckley. Killebrew and McCovey don't really fit the profile and Joe Start is all career. Palmeiro is defying comparison but I think there's no question he's behind Murray and McGwire. In fact, I'd say Palmeiro is kind of a lower ceiling version of Murray.

And Clemente! Although Clemente will likely climb when I get my updated WARP done back through the whole Retrosheet era, since his D is probably closer to +200 than +100 (arm alone is +80 according to CHONE).

One thing I've been kludging in my spreadsheet forever is comparison versus average _for position_. If could better explain to me how to determine positional average in your data (WAPA), I would probably be able to redo the whole thing in the next year to something I like better. Adding X number of wins to the SFRAC would probably get screwed up by playing time.

The main gripe about my rankings for pitchers is you leave out relief innings. Does CHONE include relief innings? If so, I might be able to swap out WARP1 for CHONE.

DL from MN, I can get those for you, as they only take a few minutes to calculate. I used to have them myself before a hotel lost a bag with my laptop in it (they're used in the stdev calculations). What baseline do you want to compare to? Within-league or both leagues, and within-season or 9-year moving average?

Yes, CHONE includes relief innings--and moreover, he does them almost right (crediting P for half their leverage above 1 to account for chaining). The only thing he's missing is the higher RP replacement level--he seems to use the same one for SP and RP. I've always said my pitcher WARP were a work in progress (and now that I've lost the spreadsheet, I'll have to redo them from scratch). By the next time we consider a backlog P (2012), all the kinks should be worked out. :)

Isn't accounting for chaining AND higher relief replacement level double counting?

Also, relief replacement level has not been constant over time. Relief Ace replacement level was a lot lower than 'closer' replacement level.

But if you are accounting for chaining I don't believe you should be comparing to just closers or relief aces, but all relievers in one pool (which has a much lower replacement level, obviously).

No, it's not double counting. And of course you compare to all relievers in one pool, that's how chaining works...everyone moves up a slot in the bullpen until the replacement pitcher in a reliever role (.470 wpct per Tango) becomes the new mopup man. There's no such thing as a "replacement closer" or "replacement setup man."

Yes, the historical gap b/w SP and RP replacement levels is a subject that needs to be studied. Note that as RP rep level moves down over time, SP rep level has to move up to compensate, otherwise you're just saying that pitching in general in one era was more valuable than in another (which may itself be true as the Three True Outcomes represent an ever-greater percentage of PA, but that's a separate kettle of fish).

Both league 9 year moving average would be most useful I think.

I have a pitcher WARP spreadsheet, I'll send it soon.

Why does SP level have to move up? Wouldn't you be saying that a good starting pitcher is better than a reliever? I would rather not treat starter and relief pitcher innings separately.

Because if one goes down and the other doesn't go up, then the overall pitching rep level goes down. To be fair to all eras, the global replacement level (i.e. how many games an all-replacement team would win) has to be constant from year to year. So if you decrease RP rep level back in time, something has to go up to compensate--either SP rep level, or the rep level of all the position players. (You'll notice that I keep the global position player rep level constant at 1.5 stdev-adjusted wins below the pitchers-included league average per 1.00 SFrac in every season).

I'll get cracking on that file for you now, DL from MN.

I still believe pitching replacement level should be the same across the board.

Starters get the benefit of more innings, relievers get the benefit of fewer, but higher quality (and for the good relievers higher leverage) innings.

Equilibrium is achieved through the innings, not separate replacement levels.

But...but...that's empirically false, Joe. Replacement pitchers, defined however you define them, perform substantially better as relievers than starters--according to Tangotiger, replacement pitchers have a .470 winning percentage as RP and a .380 winning percentage as SP. Pretending otherwise because you find it more mathematically convenient isn't an option if we're looking for an honest assessment of value.

But that is precisely because they are throwing fewer innings.

Not to mention that the only pitchers that end up in the pool you can study that way are borderline starters anyway. The great relievers generally come up as relievers any more and only relieve. Likewise for the great starters.

It's why Brad Penny can throw 5 MPH faster in the all-star game.

Managers make the conscious decision to trade innings for quality when making pitchers starters vs. relievers.

When ranking individual pitchers, relievers lose out in the comparison if you dock them for both lack of innings AND higher replacement level.

I see it as similar to hitters hit as hitters. Not as shortstops or first basemen. You then add in their defensive contribution, which is obviously higher for SS than 1B.

Pitchers don't pitch as starters or relievers, they pitch as pitchers.

If you want to then adjust for innings, leverage, and chaining I understand. If you want adjust for inherited and bequeathed runners, that's fine (and you'll pick up some of that difference between starters and relievers when you do this).


But you lose me when you say that relievers have a higher replacement level.

Maybe I'm missing something here.

This line jumped out at me in post #55

Taylor, Ben.. 37.0 0.3 8.0 60.0 (estimated)

By those numbers, Ben Taylor is in the neighborhood of Bob Johnson and Gavvy Cravath, and ahead of inducted first basemen Keith Hernandez and Will Clark. Cravath and Clark both have a peak advantage over Taylor but Hernandez is a very good comparison.

To be fair to all eras, the global replacement level (i.e. how many games an all-replacement team would win) has to be constant from year to year.

Why?

It seems to me that the freely available talent level is likely to have gone up and down over time. In the 1870s many of the best players were not in organized ball, pushing replacement level up. In the 1890s, the explosion in "major" leagues likely drove it back down. 1930s with the huge numbers of minor league clubs, there are likely more of the Ken Phelps all stars then in the 1970s when there were many, many fewer minor leagues. Then in the 90s and 00s, it seems to have trended up with the growth of independent leagues, plus deeper college ball, plus growth of international ball. Heck, your own standard deviations should show a shifting replacement level. If there is a smaller SD, then replacement is closer to average than otherwise.

Ignoring timelining and comparing only against the existing league norms, I see no logical reason to assume that the replacement level is the same in 2009 than it was in 1989, 1959 or 1919.

Did you see my ballot? I voted for Ben Taylor, he was in the neighborhood of Bob Johnson and Gavy Cravath. It's my estimate. I think Hernandez is a good comparison, especially for fielding. Taylor had a terrific reputation for scooping throws when gloves were just... gloves.

Joe Dimino--OK, let's do a work-through example of Mariano Rivera this year. I'm using WPA rather than raw runs allowed because it's what I have handy, but the principles are the same. Here is what the average bullpen looked like from 2004-08:

Rank = rank on team in leverage index
LI = leverage index
WPA = win probability added
WPA/60 = win probability added per 60 IP
WPA/LI/60 = WPA/60 divided by leverage index, or the WPA the reliever would have if he pitched 60 innings of an average impact on the game's
outcome



Now what happens if we put Mo in this pen? He threw 66.1 innings at a 1.72 LI and posted a 3.99 WPA. He threw more innings than the average closer with a lower LI. If we transfer those over to the setup man, we get the following:



Conversely, what happens if there is no Mo, or any other #1 reliever for that matter? Well, we go back to chart 1, and move everyone up a role. For the 6+ guy, we insert a replacement pitcher. According to Tom Tango, a replacement pitcher used as a reliever will have a .470 winning percentage, which translates to a WPA/LI/60 of -.20. Thus, the replacement pen looks like this:



Rivera's value above replacement is thus the gap between bullpen 2 (5.36 WPA) and bullpen 3 (1.65 WPA), or 3.71 wins.

Do you disagree with any of this methodology?

It seems to me a reasonable shortcut to approximating closer value is simply to multiply their LI by their wins above *average* rather than replacement. In Rivera's case, that would be ((66.3*4.82/9)-14)*1.72 /10 = 3.70. Not bad! :)

dlf, well, if you don't have the same overall replacement level every year, doesn't that violate "a pennant is a pennant?" You'd be giving more wins for a pennant in a year with a low rep level than with a high one. My understanding is that's unconstitutional in the HoM (after you account for league strength and such).

Using that quick-and-dirty method on Rivera's whole career minus 1995 would put him at 52.6 wins above replacement in regular season play. I buy that.

Dan, I will read that over tomorrow, when I'm not exhausted and my brain can work. Thanks for taking the time . . .

I voted for Ben Taylor too guys! Had him 9th!

What I've been sorta doing is breaking them down into groups by the years they played, and looking at who stands out in each group and where the list sort of 'floods', i.e. where does the 'big bunching' start? Then cut it off just above that. It's an eyeball process.

Well, that's what I'd do, too. You 'only' need pitchers with ten-year careers to start with. But I might try and group them by seasons, and then work out the average career-length for pitchers in that season, and then do the standard deviation of the average career-length for all seasons, and see if there were any jumps. That might reveal a 'period' structure.

But this might be what Dan R is saying creates problems in the 1940s and 1980s. Is it therefore 'harder to pitch'? Do those pitchers need a bonus? But once we know where the anomalies lie, we might be able to dig deeper via non-stat methods, especially if it's only 10-15 pitchers.

[sigh]... I have too much marking to do to get distracted this weekend!

dlf, well, if you don't have the same overall replacement level every year, doesn't that violate "a pennant is a pennant?" You'd be giving more wins for a pennant in a year with a low rep level than with a high one. My understanding is that's unconstitutional in the HoM (after you account for league strength and such).

I don't see that. Total wins definitionally equals WAR + Replacement, right? Grossly oversimplifying, if Billy Joe creates 100 runs where a replacement player would have created 50, he is worth ~5 WAR, but if he creates those same 100 runs where replacement is 70, he is worth ~3 WAR. There would be more WAR in a low replacement level league, but not more wins. I'm only a kibitzer here and not a voter, but I don't see anything constitutionally improper in having the number of available WAR move up and down as the replacement level does. The only way replacement can stay fixed is if you assume replacement is you or me or some random guy off the street rather than the best of the freely available talent.

I think if replacement value fluctuates it can't be much more than tenths of a win. It is calculated by freely available talent which assumes there are > 1 players at replacement level to pick from. On a bell curve and with only 40 roster slots available that's a good assumption.

The whole concept of replacement or FAT has always bothered me. I mean, I don't have a proposal of how to live without it. You know, it's like women.

But seriously. I agree that the POOL of FAT is probably very constant except for very rare events such as integration, expansion, etc. But the actual performance of selected members of that pool, it seems to me, would be subject to wild swings depending on how those individuals are used by their teams, whether they' injured, whether they slump in limited playing time, etc.

Can somebody address that? I mean, how many players make up FAT and why are their numbers not subject to variability other than which is attributable to the pool as a whole?

It would be valuable for someone to post answers to sunnyday's questions.

I would only note that dlf and Dan's disagreement seems partly to be a result of different goals. In theoretical terms, dlf is most likely correct that replacement and does vary, in an absolute sense, over time, with changing conditions in the professional sport. However, the significance of those changes to the assessment of merit is questionable.

If we want to know precisely how valuable a player was in a given year, in terms of wins above replacement level, then we should have as precise a calculation of replacement level as we can (acknowledging always that replacement level is a statistical model, not an empirical measurement).

If we want to compare the _merit_ of two players who played in different eras, this measure of value may not be a valid basis for determining merit. Since a varying replacement level is contextual, not absolute, (being a result of what the other players were like at the time each of the players in question were active in baseball), some of the apparent difference in value might be attributed to context, and not to the players' relative merit. Therefore, for the purposes of determining merit, we might want to use a more stable benchmark than year-by-year replacement levels provide. Of course, it is important to use a benchmark that does not radically falsify the nature of value. Win Shares, for example, sets its zero point so far below actual replacement level that it makes durable players look much more valuable to their teams than they actually are.

But unless replacement levels vary so much between eras that they significantly alter the balance between high durability and high quality of performance as contributing factors to player value, smoothing out the differences between individual seasons and even between eras by using a standardized formula to set "replacement level" will give values that are much more helpful to someone who is trying to establish the relative merits of players from different periods than the values that would be provided by a system that reports a player's value in terms of a contextually precise replacement level.

I had a chance to ripple Dan R's WAPA2 numbers into my spreadsheet. If I go with the numbers there are some interesting consequences:

Good players get better, average players get worse (this is what I wanted to see)
1B and SS go up
3B go down - I may have to add a 3B bonus
2B, OF, C stay the same (modified the C bonus to get this result)

Here's a modified prelim ballot based on the modification of my "positional average" kludge to something measured:

1) Bagwell - no real change in numbers
2) Brown
3) Walker - no change
4) Cone
5) Bridges
6) Tiant
7) Palmiero - +2 pts
8) Reuschel
9) Shocker
10) Ben Taylor
11) Phil Rizzuto - +6 pts, on ballot now
12) Kevin Appier
13) Gavy Cravath
14) Bob Johnson
15) Bert Campaneris - +8 pts, on ballot now

16) Bus Clarkson
17) Dave Bancroft +++
18) Norm Cash
19) Johnny Pesky ++
20) Dick Redding
21-25) Wally Schang, Lave Cross, Dave Concepcion ++, Tony Mullane, Jack Fournier +
26-30) Dizzy Dean, Tommy Leach --, Dutch Leonard, Jim McCormick, Virgil Trucks
31-35) Fred Dunlap, Babe Adams, Leroy Matlock, Ron Guidry, Vern Stephens +

Like I said before, 3B drop a LOT. I may need to adjust them back up some. Mike Schmidt dropped out of the top 30 all-time which doesn't smell right.

DanR - I suppose that does seem like a pretty reasonable way to handle it.

How would you handle the value of a setup man? How do you decide if your guy is a #2, #3, #4, etc.?

And the bigger question - how do we figure this out throughout history? Obviously the typical bullpen of 1960 vs. 1980 vs. 2009 are very different animals.

Also, I don't want to use WPA at all - it has some major issues, even though I do think it's reasonable for relievers, I'd much rather use runs allowed. Although I guess I could convert the WPA wins to runs if I had to, then adjust for team defense (since WPA doesn't). One advantage would be I wouldn't have to use the Inherited/Bequeathed runners stats. I guess I'm open to persuasion either way.

And, I still think that replacement level is too high.

I mean if the back end of an average bullpen is -.07, I find it tough to swallow that there are free guys out there at just -.20.

sunnyday2: Well, I use a nine-year moving average of the worst 3/8 of major league regulars. At its worst (1893, where I don't have pitcher's box era data so I am only looking at the 1893-97 period), that's 5*12*3/8 = 22-23 players per position. By its best (the modern game), we're looking at 9*30*3/8 = over 100 players. The completely unproven and untested assumption is that the random variation in FAT player usage you describe washes out with the sample size.

DL from MN--I warned you that wins above positional average is susceptible to star droughts and gluts! The 70s and early 80s were the golden age for 3B. You had Bando, Bell, Boggs, Brett, Cey, Darrell Evans, some Harrah, some Killebrew, Bill Madlock, some Molitor, Nettles, the late Brooks, a bunch of Rose, the late Santo, and the late Torre, in addition to Schmidt. Of course wins above positional average will underrate him! I don't know why you'd prefer it to plain old WARP that have none of these issues...but, to each one's own. I imagine earlier guys like McGraw, Jimmy Collins, and Baker would come out allright, no?

Joe Dimino:

1. You just slot a setup man into the #2 role and proceed as in the prior example.
2. You decide if a guy is #2/#3/#4 by leverage index. (If you did it right, you'd have a smooth curve of LI rather than distinct roles--I could probably do the math to get that if I took half an hour).
3. You figure this out throughout history with a cursory study of the evolution of bullpen usage. I'd do it myself, but I don't know how to do Retrosheet queries. (Pre-Retrosheet would be tougher, but there are no serious RP candidates from that era, making it irrelevant for our purposes).
4. You absolutely do not need to use WPA. Just replace every reference in my prior post to WPA with "runs above average" and it works just as well.
5. Why does that seem high to you? Remember, the worst 3/8 of starting MLB position players tend to track replacement level pretty well. Here, we're talking about the bottom 1/5 or so of relievers, and then subtracting .13 context-neutral wins per 60 innings from *that*. If anything, that seems generous, no? In the end, of course, your beef is with Tom Tango, the source of the .470 figure. You should write him if you'd like to, well, tango on this issue.

If I'm a GM and there are a ton of average 3B, then there's probably not much incentive for me to give up much for one in trade. If talent is top heavy then there isn't that as much advantage upgrading at that position. There's the wins required to get to average (not hard to get) and the wins to go from average to contender (really hard to get).

The earlier guys do turn out fine, and the 70s stars didn't lose much after I corrected for DH time, it's the average guys that get hurt. I like to see that separation.

When you say "corrected for DH time," you mean "gave credit for seasons primarily played at DH," right? The DH adjustment is already baked into these wins-above-positional-average figures.

Yes

The Rosenheck DB will be up shortly (FTP client says 17 minutes). It's the older version though, but I'm not sure when I'll get to update it for his latest baserunning fielding adjustments (some time this weekend or possibly not until early next week) and didn't want to wait.

This is a total digression from where this thread has gone, but I just had the idea early today and it still seems good to me, and this thread is active right now, so:

How about we take some time and fold the newly-elected guys into the positionals we just spent a year doing? In other words, where do Larkin, Alomar and Edgar rank within the already-elected guys at shortstop, second base, and wherever Edgar ends up? It seems that this should not take too long, as the new guys are fresh in our minds, and it saves having to revote the whole thing in five years or so. It also brings up the hilarious topic of what list to put Edgar on - whether to start a new DH "position" or just fold him into the first basemen or something (third base would be a weirder choice than first, given Edgar's actual achievements). I, personally, would much rather do this now than in five years with 15 guys to fold in, but I'm hardly the only one involved. - Brock

How about we take some time and fold the newly-elected guys into the positionals we just spent a year doing? In other words, where do Larkin, Alomar and Edgar rank within the already-elected guys at shortstop, second base, and wherever Edgar ends up?

That oughta have it's own thread or maybe be on the current results thread? But I'll bite anyway, at least Larkin for now. Bill James has him #6 all-time, for what it's worth, which seems high. Not to mention at #6 Larkin is behind Banks and ahead of Davis and Dahlen, and I'd be pretty sure we would reverse that.

I don't fully agree with our consensus result at SS of course, but taking the consensus nevertheless, I'd plug Larkin in thusly:

Inner Circle

1. Wagner
2. Lloyd
3. Ripken

Easily Qualified

4. Vaughan
5. G. Davis
6. Dahlen
7. Yount
8. Appling
9. G. Wright
10. Cronin

Qualified Even for a Small Hall

(Larkin?)
11. Banks
(or, Larkin?)
12. Wells
13. Johnson
14. Ozzie

It Takes a Large Hall

15-19. Trammell, Reese, Boudreau, Glasscock, Ward,

These Could Be Debated Even in a Large Hall

20++.Jennings, Pearce, Wallace, D. Moore, Lundy, Sewell

Larkin is clearly above the bottom 2 groups. Clearly better than Ozzie. Comparable to Banks? Well, not really for reasons of style. Probalby most comparable to Wells and Johnson. Better than Cronin? No. So maybe 11th ahead of Banks, or 12th just behind Banks. I understand I probably have him lower than many though he was #1 on my 2010 ballot.

As to Edgar, I think you have to rate him as a "hitter," which means among the 1B-LF-RFers, all of them, whichh of course means integrating the three positions, something we haven't done yet or ever.

Replacement Level as a tool can remain constant, if that is what is appropriate to whatever metric you are using it for.

Replacement Level as a reflection of reality should change based on the quality of players available. I'm not saying that is easy to do, but it is certainly more accurate.

In fact, TRUE Replacement Level could vary from team to team, though that would really throw valuation out of whack (though it may give a 'better' interpretation of 'valuable').

Sunnyday -

I really appreciate the support, but I just realized, looking at my own post, that I didn't make it clear that I did not intend to HIJACK THIS thread. If there is support, of course the work should go into a separate thread. I only posted up here because this is the thread that HoM people are still checking in some numbers. If there is any sign of interest, this should go into its own thread, even to finish the discussion of whether to do it or not. This thread is this thread. I just don't have the permissions to start one of my own. Joe! Help! - Brock

That could work brock although if we get the MVP project off the ground we may be busy enough . . . regarding Edgar's position on the list . . . since it doesn't make sense to rank him as a DH (he's #1 so far), I would say he'd be on the 3B list. He played 563 games there and 28 at 1B.

You would just treat him like a Joe Torre on the catcher's list - rank his career value and slot him where he'd fit, considering him with much less defensive value than the other 3B on the list.

My guess is that he'd end up ahead of Nettles and Boyer, below Hack.

3B results

One way to do it might be to require everyone to vote the same ballot they already voted, but slot Larkin, Alomar and Martinez accordingly. So we maintain some consistency.

One way to do it might be to require everyone to vote the same ballot they already voted, but slot Larkin, Alomar and Martinez accordingly. So we maintain some consistency.

No new voters?

Joe, I think you made my point for me. Rating Edgar at 3B would seem to be an impossible job. Ranking him as a hitter I think I could do.

Thx for the plug and the reminder. Coming soon to a Webblog near you: The MVP Project!!!

88. bjhanke Posted: December 04, 2009 at 05:01 AM (#3403138)
How about we take some time and fold the newly-elected guys into the positionals we just spent a year doing? ... I, personally, would much rather do this now than in five years with 15 guys to fold in, but I'm hardly the only one involved. - Brock

Already there are six,
Henderson, McGraw, Smith
Larkin, Alomar, Martinez

Ooops, I was correcting the spreadsheet today and noticed that Palmiero actually drops just off the end of my ballot. I think the characterization of him as a lower ceiling Eddie Murray is still accurate. He's above the standard but much more of a borderliner than I thought previously. Jake Beckley took forever, Palmeiro shouldn't go in that smoothly.

I disagree, DL from MN. 66 WARP2 is just too much career to ignore. You'd have to be a 100% pure peak voter to dismiss Palmeiro.

I didn't even come close to getting a ballot ready for 2010, but I should have plenty of time for this one. I'm going to write an application to pull my numbers together at some point - doing it by hand takes too long (I'll be happy to share the source code for that when I'm done if anyone happens to be interested).

I look for a blend of peak/prime. The peak is the primary measure, but I believe a meritorious player ought to be able to contribute meaningfully at the major league level for a significant (10+ years) period of time. The peak itself need not be that long, of course, but the length of contributing play is sort of a minimum barrier that must be cleared. So for instance I'm not high at all on, say, Dizzy Dean because I see his entire career as being 6 years.

This is an incomplete list of some of the players from last year's ballot along with one newcomer, Bagwell (my first guess is that he'll be a clear #1 for me). This is the AROM WAR-based list - eventually the list will be a composite system. I'd just like to get started on explaining my methodology so folks can critique it. Note this is not the top of the ballot necessarily - just where these particular players rank relative to one another (other players may slot in between these).

what you see is:

name
avg war / yr over what I consider the span of contributing play,
length of span (longest consecutive number of years within that span),
seasons at each WAR benchmark

jeff bagwell
5.67 WAR / yr, 14 yrs (14 cons.), 9+ 1, 8+ 3, 7+ 4, 6+ 5, 5+ 9, 4+ 11
massive peak over a very long period of time with a substantial amount of wins even in non-peak seasons - this is exactly the sort of career shape that is ideal for doing well in my system (it's a ground-breaking conclusion, I know)

rick reuschel
4.45 WAR / yr, 15 yrs (9 cons.), 8+ 1, 5+ 7, 4+ 8
best 'best year' out of the next couple of guys and the longest period of sustained peak play, plus highest average rate on the list after bagwell (a minor factor, really, but meaningful since all the others are close to one another)
this is an example of this project opening my eyes to a player where otherwise I'd have had no idea he was this good

luis tiant
3.80 WAR / yr, 16 yrs (9 cons.), 7+ 1, 6+ 2, 5+ 6, 4+ 7
6 years at 5 WAR or better is a lot

dale murphy
3.62 WAR / yr, 13 yrs (13 cons.), 7+ 3, 6+ 4, 5+ 6
probably the best peak on here after bagwell but...absolutely no depth whatsoever
I may have him too high - his 'contributory play' years are mostly just a tick above replacement level and possibly I should be discounting him for a too-short career

david cone
3.99 WAR / yr, 14 yrs (13 cons.), 6+ 3, 5+ 4, 4+ 7
peak not incredibly high but kept it up for longer than anyone else below him here

gavy cravath
3.59 WAR / yr, 9 yrs (8 cons.), 7+ 1, 5+ 2, 4+ 4
he's being given extra credit to rank here
I doubt I'll be confident on his ranking until I move to a consensus-based analysis
AROM gives him only one really eye-catching season and I keep going back and forth between thinking I'm giving him too much extra credit and not enough - he had a historic-level bat

bucky walters
3.85 WAR / yr, 12 yrs, (12 cons.), 9+ 1, 6+ 3, 5+ 4, 4+ 5
massive best year but peak was short
I have him a tick behind Cone but they could easily switch places

phil rizzuto
3.73 WAR / yr, 14* yrs (14* cons.), 7+ 1, 5+ 2, 4+ 5
even with war credit he doesn't do all that well - I'm hesitant to credit 43-45 at a level substantially above the average I have here

tommy leach
3.39 WAR / yr, 15 yrs (14 cons.), 6+ 2, 4+ 5
AROM doesn't like him at all - 2 really strong years but without a whole lot of supporting cast

You decide if a guy is #2/#3/#4 by leverage index.

That doesn't work for anything prior to about 1995, because relief roles other than the end-of-game reliever weren't that stratified. You didn't start seeing true setup men until about 1990 (Mike Jackson and Duane Ward). The teams that had multiple-reliever bullpens, like the A's, Phillies and Pirates of the 70s, did so much mixing and matching that you can't really tell who was doing what other than the closer. Before the 90s, you pretty much had the closer and then everyone else, and anyone else in the pen could and did pitch in any role.

-- MWE

The method was designed for the modern game. Applying it historically would require a study of the evolution of bullpen usage over time, which I'd be happy to do if someone could put in the Retrosheet queries for me.